Adaptive Stochastic Gradient Descent for Fast and Communication-Efficient Distributed Learning
This work addresses the issue of slow workers in distributed machine learning, offering an incremental improvement for faster and more communication-efficient training.
The paper tackles the problem of stragglers in distributed stochastic gradient descent by proposing an adaptive policy for selecting the number of workers to wait for at each iteration, optimizing the error-runtime trade-off. The results show that this adaptive version achieves lower error values in less time and reduces communication compared to non-adaptive implementations.
We consider the setting where a master wants to run a distributed stochastic gradient descent (SGD) algorithm on $n$ workers, each having a subset of the data. Distributed SGD may suffer from the effect of stragglers, i.e., slow or unresponsive workers who cause delays. One solution studied in the literature is to wait at each iteration for the responses of the fastest $k<n$ workers before updating the model, where $k$ is a fixed parameter. The choice of the value of $k$ presents a trade-off between the runtime (i.e., convergence rate) of SGD and the error of the model. Towards optimizing the error-runtime trade-off, we investigate distributed SGD with adaptive~$k$, i.e., varying $k$ throughout the runtime of the algorithm. We first design an adaptive policy for varying $k$ that optimizes this trade-off based on an upper bound on the error as a function of the wall-clock time that we derive. Then, we propose and implement an algorithm for adaptive distributed SGD that is based on a statistical heuristic. Our results show that the adaptive version of distributed SGD can reach lower error values in less time compared to non-adaptive implementations. Moreover, the results also show that the adaptive version is communication-efficient, where the amount of communication required between the master and the workers is less than that of non-adaptive versions.